RESULTADOS

Description: A network failure, caused by e.g. failing equipment or a cyberattack, will influence the algorithm. The auctioneer does not receive all information in this case and will therefore base its clearing price on only a subset of the demand. In order to investigate the impact on the power consumption, a house with load devices, solar panels and a battery, which were all connected to an auctioneer, are emulated. During the experiment, the communication between the clients and auc- tioneer will be disconnected.

Experiment: Measurement data of a real house is used for the desired power consumption of the devices and the production of energy from solar panels. The battery that is modeled is a Tesla Pow- erwall 2, with a maximum capacity of 13.5 kWh and a maximum (dis)charge rate of 5 kW [43]. The auctioneer tries to clear the market at a total power consumption of 0 W.

The devices form one demand function together, which consists of a small uncontrollable section and a larger flexible part. They only accept their total desired power consumption for very low prices. The solar panels are willing to offer all power they produced, except for a small range of low prices, in which curtailment will be applied. The battery wants to charge when the price is low, discharge when it is high and takes no action for prices in between. It depends on the state of charge of the battery at which rate it demands to charge or discharge. It starts at a state of charge of 0 Wh. Clients that are disconnected from the auctioneer will consume or produce the amount of power that they would do without control. The auctioneer clears the market once every 5 minutes, although in the emulation this is set to 5 seconds, in order to speed up the process.

and the load devices was lost. The devices were back online at interval 161. Then at interval 175, the solar panels are disconnected and from interval 203, they were connected again.

Result: Figure 5.9ashows the emulation without disconnections. In Figure 5.9b, the disconnection and reconnection of the load devices and the solar panels, respectively, is shown. Only the interesting part from interval 100 until 270 is shown. Outside these intervals, the power consumption is roughly zero.

Inb, it can be seen that as soon as the load devices are disconnected, the battery charges following the offered power by the solar panels. Therefore, the netto power consumption follows the consumption by the load devices. Hence, this power must come from other sources. The battery also does not compensate for the peaks around interval 150.

During the time that the solar panels are disconnected, according to the auctioneer, the battery has to provide all the needed energy. Since the capacity of the battery is limited, the auctioneer tries to restrict the power usage of the devices in this situation.

It can be seen that in both cases, the total power consumption drops back to zero soon after the connection is established again. However, the state of charge of the battery will be different, because it did not discharge while the load devices were offline and it could not charge while the solar panels were disconnected. −1000 −500 0 500 1000 a P ow er [W] Load PV Battery Total 100 120 140 160 180 200 220 240 260 −1000 −500 0 500 1000 b Interval P ow er [W]

Figure 5.9: a)Power consumption of the load devices, PV panels, battery and their sum with normal control. b) Power consumption of the clients and their sum during a network error at the load from interval 133-161 and at the PV panels from 175-203.

Discussion: During the disconnected phases, the total power consumption deviates from the de- sired 0 W by the amount that the disconnected clients consume or produce. This misbehavior might even be less favorable than no control and should be avoided. The auctioneer can find out when a device is suddenly disconnected, but it is difficult to say if this is on purpose or not. In order to imple- ment error handling, a client and auctioneer should come to an agreement in case the client wants to close the connection. Then there are several options to undertake when a real communicational error

occurs. If electrical measurement equipment is installed in the relevant building, this can be used by the HEMS to calculate a local market clearing, as described in [1]. Nonetheless, this requires the HAN to work, such that the clients can still be controlled. If this is not the case, a client can also consume or produce a certain amount based on the latest prices it received or on a pattern from earlier similar periods. This can work decently for a short period, but cannot be used on the long run, because the system is too dynamic.

Chapter 6

Recommendations

In this chapter, a perspective and recommendations for the choices to be made regarding communi- cation in smart grids, especially for the deployment of the active control methodology, are presented. Next to that, discussion points on the conducted research and several suggestions for future work are given.

6.1 Network technologies and layout

Inside buildings, most likely wireless technologies, such as Z-Wave and ZigBee, are to be deployed in order to let devices communicate with the HEMS. As mentioned in Section 2.3.2, several reasons exist why connecting smart meters in the NAN should be done using a Wireless Mesh Network. Standards such as IEEE 802.11s and IEEE 802.15.4g can be used for this. In order to extent the transmission range for less dense neighborhoods or in order to connect multiple NANs, other technologies such as Long-Term Evolution (LTE) or its successor 5G will be needed. A hierarchal tree structure is more suited on this scale. IEEE 802.11ah might be appropriate to deliver data up to the WAN, but needs further research before it can be realized.